This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Perspectives and Limits of Optical Information Processing Kelvin Wagner University of Colorado, Boulder CO 803090425 [email protected] http://optics.colorado.edu QELS Quantum Information Science Symposium – May 9, 2001 Branches of Optical Information Processing Optical Signal Processing Digital Optical Computing Optical Interconnections Optical Memory Optical Neural Networks Algorithms for Optical Processors Physics of Optical Materials and Devices Fundamental Limits and the competition Applications, Packaging, and Commercialization Kelvin Wagner Applied Optics special issues on Optical Computing: May 1 88, May 10 90, Sept 10 92, Mar 10 1994, Mar 10 96, May 10 98, Feb 10 2000 Why Optical Computing? • Massive Parallelism Arrays of 10 3 × 10 3 devices and SBWP of 10 4 × 10 4 • Huge Bandwidths Capabilities 10 fsec pulses imply 100THz BW, but IO device limited • Global Interconnectivity without wiring penalty Optical beams cross through each other without interacting • Physics is matched to certain problem Why not? • Only weak nonlinearities require high power • Special purpose vs general purpose computation Kelvin Wagner 1 T.E. Bell, Optical Computing: a field in flux, IEEE Spectrum, p 34, Aug 1986. Byte September 1992, pp 2275. IEEE Computer Feb 1998. Proc IEEE special issues on Optical Computing, July 1984, Nov 1994. Early Optical Calculator Lehmer’s Photoelectric Sieve for Prime Numbers D.H. Lehmer, A Photoelectric Number Sieve, Amer. Math. Monthly, 40, 401, 1933 • Gears with a hole allowed a light beam to pass • Residue arithmetic: Relatively prime number of teeth on gears • Factored Mersenne number 2 79 1 • Rapidly sifted out non factors (2 × 10 7 /hr) • Thousands of times faster than available tech niques at the time • in 2000 Twinkle proposed by A. Shamir to in crease factorable number size to 200 bits. a a A. Shamir, Factoring large numbers with the TWINKLE device, Weizman Inst. preprint. Kelvin Wagner I 1 Coherent Optical Processing A lens takes a Fourier transform of the coherent field at its rear focal plane θ λ θ F ( u , v ) = F x λ F , y λ F ¶ = Z Z f ( x , y ) e i 2 π λ F ( xx + yy ) dxdy 2D Processing Power: 10 3 × 10 3 FT in 1 nsec (F=15cm) ( 10 6 ) 2 ⊗ / 10 9 s = 10 21 analog multiplies/sec limited by CCD detector array readout, and SLM frame rate ( 10 6 ) 2 ⊗ / 10 2 s = 10 14 analog multiplies/sec FFT equivalent – N log N : 20 × 10 6 ⊗ / 10 2 s = 2 GFLOP Kelvin Wagner II 1 K. Preston, Use of the Fourier Transform properties of lenses for signal spectrum analysis, in Opt. and EO Information Processing, J. T. Tippet ed, MIT press, 1965 Fourier Optics, J. W. Goodman, McgrawHill, 1996 Optical Synthetic Aperture Radar Processor f 4 d 34 + + f 5 Optical Axis Multipleaperture Fourier filter Lens 1  Spherical Lens 2  Spherical Lens 3  Cylindrical Lens 4  Cylindrical Lens 5 Cylindrical Output image plane Real range focal plane f s f s d f r d 23 + + f s f 3 d 34 + +...
View
Full
Document
This note was uploaded on 12/27/2011 for the course CMPSC 290h taught by Professor Chong during the Fall '09 term at UCSB.
 Fall '09
 Chong

Click to edit the document details